MilFi

MilFi turns Ukraine's "points-for-kills" air-defense incentive system into an agent-verified, instantly-settled on-chain network.

We are a private counter-UAV contractor in Ukraine: power plants and government facilities hire us, our crews, and our interceptor drones to defend them against Shahed-type attack drones. Ukraine's government rewards confirmed intercepts with e-points that units spend on a state-run weapons marketplace. The incentive works — but confirmation is manual and settlement takes months. A unit that downs a drone tonight re-equips next quarter.

MilFi replaces that pipeline end to end:

Report — a civilian spotter, cleared by a World ID Selfie Check, photographs the threat. Agent A — a vision model running TEE-sealed on 0G Compute returns a strict JSON verdict on whether it's a real threat, produced in an enclave outside the control of anyone who profits from the payout. Intercept — a document-verified (World Identity Check) military crew engages it. Agent B — a second 0G agent judges the post-strike imagery: destroyed or not, and whether it's consistent with what Agent A saw. Two independent models on two images is much harder to fake than one frame. Evidence — every image hash and verdict is published to a Hedera Consensus Service topic. Only SHA-256 hashes go on-chain; the footage never leaves the platform, so the trail is publicly auditable without leaking a pixel of military data. Settlement — an autonomous Hedera Agent Kit agent reads that topic through Mirror Node, applies the payout rule set by Orb-verified administrators, and transfers DEFPOINT (HTS) to the unit's Hedera account itself. Seconds, not months. Before any money moves, the settle-agent must obtain a payout authorization from World AgentKit. Our claim-agent is registered in AgentBook on World Chain and resolves to a real human's World ID nullifier. A bot with no human behind it gets a 402 challenge it can't answer, an unregistered signer gets 403 — and the payout is refused before a transaction is ever attempted.

It runs on top of the air-defense coordination map we already operate in production.

Live on Hedera testnet and independently checkable: DEFPOINT token 0.0.9753000, evidence topic 0.0.9753001, real autonomous payouts on HashScan. Zero Solidity anywhere — HTS, HCS, Scheduled Transactions and Mirror Node only.

Live demo: https://mil-fi.onrender.com — sign in as government, military, or spotter, password milfi.

Different World ID tiers gate different roles for a reason: a Selfie Check is what lets a civilian spotter's report enter the pipeline at all, and only Orb-verified administrators can set the payout thresholds the settle-agent enforces later on. An interceptor then fires from the air; the strike footage feeds directly into a 0G Compute vision model running in a TEE trust mode, so the threat/kill verdict is produced somewhere neither we nor the paid party can influence, and military data stays safe. That verdict, and a hash of the footage, get posted to a Hedera Consensus Service topic. A Hedera Agent Kit agent reads that topic through Mirror Node on its own schedule, decides the conditions are met, and moves DEFPOINT tokens on the Hedera Token Service straight to the unit's account. It only fires the transfer after checking, via World's AgentKit, that the wallet asking to be paid is actually backed by a real, unique, verified person; a bot with no human behind it gets refused.